severe local storms forecasting

27
WEATHER AND FORECASTING VOLUME 7 Severe Local Storms Forecasting * ROBERT H. JOHNS National Severe Storms Forecast Center, Kansas City, Missouri CHARLES A. DOSWELL III National Severe Storms Laboratory, Norman, Oklahoma (Manuscript received 11 May 1992, in final form 13 August 1992) ABSTRACT Knowledge of severe local storms has been increasing rapidly in recent years as a result of both observational studies and numerical modeling experiments. This paper reviews that knowledge as it relates to development of new applications for forecasting of severe local storms. Many of these new applications are based on physical understanding of processes taking place on the storm scale and thus allow forecasters to become less dependent on empirical relationships. Refinements in pattern recognition and severe weather climatology continue to be of value to the operational severe local storms forecasters, however. Current methodology for forecasting severe local storms at the National Severe Storms Forecast Center is described. Operational uses of new forecast applications, new “real-time” data sources (such as wind profilers and Doppler radars), and improved numerical model products are discussed. 1. Introduction Convective storms produce a wide variety of weather phenomena that might be considered “severe” (a hazard to life and property). For purposes of this discussion, however, only those convectively induced phenomena forecast by the Severe Local Storms Unit (SELS) of the National Severe Storms Forecast Center (NSSFC) will be considered. These are: (a) tornadoes (b) damaging winds, or gusts 26 m s -1 (50 kt) (c) hail diameter 1.9 cm (3/4 inch) Doswell et al. (1993) have described present-day forecasting of tornadoes as consisting of two parts: anticipation of tornadic potential in the storm environment, and recognition of tornadic storms once they develop. This two-part forecasting/observation process also applies to damaging winds and hail. This paper considers only the Corresponding author address: Robert H. Johns, NOAA/NWS Na- tional Severe Storms Forecast Center, Room 1728 Federal Office Build- ing, 601 East 12th Street, Kansas City, MO 64106. * A preliminary version of this paper appeared in the preprints of the Symposium on Weather Forecasting, held in conjunction with the 72nd Annual Meeting of the American Meteorological Society, Atlanta, Geor- gia, in January 1992. first part of this forecasting process, focusing on the relationship between the severe local storm and its environment. Since the primary forecast product of SELS is the severe weather watch (see Ostby 1993 for a complete description of SELS forecast products), that is the primary topic within this paper. Other SELS forecast products will be mentioned as well, however. The forecasting process utilized by SELS involves parameter evaluation, pattern recognition, and climatology (see section 3a in Doswell et al. 1993). The parameter evaluation and, to a lesser extent, pattern recognition components of severe weather forecasting continue to change as more is learned about storm- scale processes and interactions between the storm-scale and the larger- scale environment. Also, climatological knowledge concerning severe local storms is likely to become more refined as specific studies are conducted to develop regional severe weather (e.g., Hirt 1985; Anthony 1988) and parameter-specific (e.g., Lanicci and Warner 1991) climatologies. To evaluate parameters and detect patterns, both synoptic-scale and mesoscale analysis are essential tools for severe local storms forecasters. The reader is referred to Doswell (1982) for a discussion of those basic analysis techniques utilized in SELS. Since the early 1980s, additional techniques have been developed to take advantage of new, advanced technology and an enhanced understanding of storm processes and their interaction with

Upload: lee-mcdermot

Post on 18-Feb-2016

245 views

Category:

Documents


0 download

DESCRIPTION

Johns and Doswell

TRANSCRIPT

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

Severe Local Storms Forecasting *

ROBERT H. JOHNS

National Severe Storms Forecast Center, Kansas City, Missouri

CHARLES A. DOSWELL III

National Severe Storms Laboratory, Norman, Oklahoma

(Manuscript received 11 May 1992, in final form 13 August 1992)

ABSTRACT

Knowledge of severe local storms has been increasing rapidly in recent years as a result of both observationalstudies and numerical modeling experiments. This paper reviews that knowledge as it relates to development ofnew applications for forecasting of severe local storms. Many of these new applications are based on physicalunderstanding of processes taking place on the storm scale and thus allow forecasters to become less dependent onempirical relationships. Refinements in pattern recognition and severe weather climatology continue to be ofvalue to the operational severe local storms forecasters, however.

Current methodology for forecasting severe local storms at the National Severe Storms Forecast Center isdescribed. Operational uses of new forecast applications, new “real-time” data sources (such as wind profilers and

Doppler radars), and improved numerical model products are discussed.

1. Introduction

Convective storms produce a wide variety of weatherphenomena that might be considered “severe” (a hazardto life and property). For purposes of this discussion,however, only those convectively induced phenomenaforecast by the Severe Local Storms Unit (SELS) of theNational Severe Storms Forecast Center (NSSFC) will beconsidered. These are:

(a) tornadoes(b) damaging winds, or gusts ≥ 26 m s-1 (50 kt)(c) hail diameter ≥ 1.9 cm (3/4 inch)

Doswell et al. (1993) have described present-dayforecasting of tornadoes as consisting of two parts:anticipation of tornadic potential in the storm environment,and recognition of tornadic storms once they develop. Thistwo-part forecasting/observation process also applies todamaging winds and hail. This paper considers only the

Corresponding author address: Robert H. Johns, NOAA/NWS Na-tional Severe Storms Forecast Center, Room 1728 Federal Office Build-ing, 601 East 12th Street, Kansas City, MO 64106.

* A preliminary version of this paper appeared in the preprints of theSymposium on Weather Forecasting, held in conjunction with the 72ndAnnual Meeting of the American Meteorological Society, Atlanta, Geor-gia, in January 1992.

first part of this forecasting process, focusing on therelationship between the severe local storm and itsenvironment. Since the primary forecast product of SELSis the severe weather watch (see Ostby 1993 for a completedescription of SELS forecast products), that is the primarytopic within this paper. Other SELS forecast products willbe mentioned as well, however.

The forecasting process utilized by SELS involvesparameter evaluation, pattern recognition, and climatology(see section 3a in Doswell et al. 1993). The parameterevaluation and, to a lesser extent, pattern recognitioncomponents of severe weather forecasting continue tochange as more is learned about storm- scale processesand interactions between the storm-scale and the larger-scale environment. Also, climatological knowledgeconcerning severe local storms is likely to become morerefined as specific studies are conducted to developregional severe weather (e.g., Hirt 1985; Anthony 1988)and parameter-specific (e.g., Lanicci and Warner 1991)climatologies.

To evaluate parameters and detect patterns, bothsynoptic-scale and mesoscale analysis are essential toolsfor severe local storms forecasters. The reader is referredto Doswell (1982) for a discussion of those basic analysistechniques utilized in SELS. Since the early 1980s,additional techniques have been developed to takeadvantage of new, advanced technology and an enhancedunderstanding of storm processes and their interaction with

DECEMBER 1992 J O H N S A N D D O S W E L L

the environment. These recent techniques and sometraditional techniques presented by Doswell will beconsidered in the following sections.

While climatological knowledge of severe local stormevents is useful in the overall evaluation of the severeweather threat, the atmosphere on any given day may notconform to the statistical norm. Severe events can and dooccur in all states, at all times of the year, and at all hoursof the day. That is one of the reasons for SELS to be inoperation 24 h per day, every day. Moreover, byconcentrating their attention on severe weather, SELSforecasters are able to provide important guidance to fieldforecasters in areas of the country where severe localstorms are infrequent. If a forecaster is not accustomed todealing with severe convective events, those events mightbe an unpleasant surprise on those relatively rare occasionswhen they do occur.

A review of the history of severe weather forecastingtechniques through the mid- 1980s has been presented bySchaefer (1986), and interested readers are referred to hiswork for a historical perspective on the subject. Theprimary objective of this paper is to review current severelocal storm forecasting techniques, with an emphasis onrecent major advances in the field. Knowledge gained fromobservational studies and numerical model simulations willbe discussed, and theories and forecast techniques derivedfrom this knowledge will be presented. The importance ofnew technology in helping to assess the potential for severelocal storm development will also be discussed.

2. Severe local storms forecasting philosophy andmethodology

SELS does not attempt to have a severe weather watch(Ostby 1992) valid for each and every severe weather event.The density of the operational data network, the state ofmeteorological understanding, and the temporal and arealscale of SELS forecasts do not allow for reasonableaccuracy in attempting to forecast all isolated, marginalsevere events. Because of this, SELS concentrates itsefforts on issuing watches for significant severe weatherevents (i.e., giant hail, strong and violent tornadoes, etc.;see Hales 1988), and concentrations of severe weatherevents [e.g.. 20 reports of 2.5-cm- (1-inch) diameter hailacross northern Missouri]. Thus, it is expected that isolated,marginal severe weather events will be handled by the localNational Weather Service (NWS) offices in the normalcourse of their warning duties.

The severe local storms forecasting methodologyutilized by SELS varies with the time scale of the forecastproduct (see Doswell et al. 1993 and Ostby 1992 fordefinitions and details). Preparation of convective outlookforecasts (out to 52 h) primarily involves interpretation

and modification of National Meteorological Center(NMC) numerical model forecast products (see section 3).For those outlook forecasts whose valid period beginswithin 12 h of the scheduled issuance time, some additionaladjustments are made based on diagnosis of currentsynoptic- and subsynoptic-scale trends.

On the 0-7-h time scale, SELS issues three types ofproducts: mesoscale discussions, severe local stormwatches (severe thunderstorm and tornado), and statusreports (Ostby 1992). These short-term products areprimarily diagnostic in nature, and if they are to be timelyand accurate, continuous attention to details and trends of“real-time” weather is required (see Doswell 1986a). Forexample, regional subsynoptic-scale analysis of surfacedata (the densest operational data network available) isnecessary to assess the short-term severe weather threat(Doswell 1982), so surface analysis typically is done eachhour. Further, these subjective surface analyses arecomplemented by parameters derived from the surface datafields. SELS mesoscale forecasters1 also examine closelyremote-sensing imagery from satellite (e.g., Scofield andPurdom 1986), radar (e.g., Burgess and Ray 1986),lightning (e.g., Lewis 1989), and wind profiler (e.g.,Leftwich and Beckman 1992) sources to aid in accuratediagnosis of important synoptic and mesoscale features.These data are available at intervals ranging from nearlycontinuous (lightning) to hourly (profiler winds).

As additional data sources have become available andNSSFC’s means of displaying data has improved, statusreports and mesoscale discussions have become moretimely and detailed in recent years (see Figs. 1 and 2). Inaddition to their original uses (see Doswell et al. 1993),these messages are now sometimes utilized to alert fieldoffices to a particularly dangerous mesoscale development(e.g., Fig. 1b), or to discuss a localized severe threat whenthe expected intensity, area affected, and / or time durationis too limited to justify watch issuance (e.g., Fig. 2b).

Although short-term forecasting of severe local stormsis primarily diagnostic in nature (Doswell 1986b), theoperational numerical weather prediction (NWP) modelforecasts also play a role. Typically, at the beginning of ashift, a SELS mesoscale forecaster examines the short-term model forecasts (6, 12, and sometimes 18 h) todetermine model trends for those parameters and patternsrelated to severe local storm development. Used togetherwith the current convective outlook, they help themesoscale forecaster focus on areas that require moredetailed analysis. Other short-term forecast products use

1 The term “mesoscale forecasters” is used to signify those SELS per-sonnel filling positions where the primary responsibility is to monitor“real-time” weather conditions and trends continuously and to issue prod-ucts concerning short-term subsynoptic-scale events when necessary (i.e.,mesoscale discussions, severe weather watches, and status reports).

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

combinations of current surface data and model data toassist the forecaster in estimating current parameter values(see section 3).

3. Initiation of deep convection

Almost all severe local storm events are associated withdeep convection. To achieve deep convection, there arethree necessary ingredients (as described in Doswell 1987):

(a) a moist layer of sufficient depth in the low ormidtroposphere,

(b) a steep enough lapse rate to allow for a substantial“positive area,” and

(c) sufficient lifting of a parcel from the moist layer toallow it to reach its level of free convection (LFC).

FIG. 1. (a) Status report issued at 0646 UTC 26 April 1991 discussingthe potential issuance of a watch adjacent to the current watch. Contrac-tion are in accordance with Federal Avaiation Admin. Handbooks No.7340.1L and 7350.6E (b) As in (a) except issued at 1945 UTC 26 April1991 and discussing a dangerous mesoscale development over north-central Kansas.

Since the essential issue in the formation of deepconvection is whether or not the LFC will be attained, thequestion of “sufficiency” in these ingredients is determinedby whether or not some parcel can be expected to becomepositively buoyant through a deep layer. Moisture,conditional instability, and lifting are all necessary and eachaffects the convective potential in a different way.

To assess the potential for deep convection, a forecastermust be able to diagnose the current thermodynamicstructure of the troposphere and to forecast changesresulting from thermal advection, moisture advection, andvertical motion fields. Currently, the diagnosis of thesefactors is done by means of twice-a- day (0000 and 1200UTC) radiosondes taken from a network of stations aroundthe world. Radiosondes give the forecaster a snapshot ofvertical thermodynamic and wind profiles at widelyscattered points. The challenge is to deduce the structurebetween observations in space and time, utilizing thelimited sounding data, and to project temporal changes inthis structure for the forecast period in question.

FIG. 2. (a) As in Fig. 1a except for mesoscale discussion issued at 1829UTC 22 July 1991 discussing synoptic and subsynoptic scale conditionsand trends that may lead to a watch issuance. (b) As in (a) except issuedat 1045 UTC 2 March 1991 and discussing a small area where the severethreat is likely to be limited in duration.

DECEMBER 1992 J O H N S A N D D O S W E L L

In the early days of sounding analysis, forecasters handplotted local radiosonde data on a thermodynamic diagram(e.g., a pseudoadiabatic chart), and assessed the potentialfor deep convection graphically, based on expected diurnalheating and other factors. Estimation of static instability(positive area) and cap (negative area) were an importantpart of this process. Since technology at that time did notallow for timely quantitative determination of positive andnegative areas, several stability indices (e.g., the SELSlifted index; Galway 1956) were developed. Such indices,many of which remain in use today, key primarily onmandatory pressure-level data. This makes them sensitiveto the details in the sounding and, hence, they may beunrepresentative of the true character of the data. Further,the sounding used in their calculation can changesignificantly in the time between the sounding andconvective development, so the numerical values can beunrepresentative in this way, as well. As technology andour understanding of storm processes have changed, newtechniques and parameters are evolving that make use ofmore of the information in a sounding than the traditionalindices. Currently at SELS, forecasters use the VAS[VISSR (Visible and Infrared Spin Scan Radiometer)Atmospheric Sounder] Data Utilization Center (VDUC)interactive computer system (Browning 1991, 1992)routinely to display thermodynamic profiles and todetermine derived parameters. Local analysis computerprograms, such as CONVECT (Stone 1988), andinteractive programs, such as the Skew T—logp/Hodograph Analysis and Research Program (SHARP; Hartand Korotky 1991), are being utilized widely by localforecasters to assist in quantitative sounding analysis.These interactive programs incorporate the old indicesalong with newer ones (e.g., convective available potentialenergy, or CAPE) and allow the forecaster to modify thesounding data to attempt to account for anticipated spatialand temporal variations in the storm environment.

To assist the SELS forecaster in defining (in both areaand time) the short-term potential (0–7 h) for thunderstormdevelopment, hourly surface data are used to computeparameter fields, some of which are modified for expectedchanges aloft.2 By this means, the forecaster can obtainhourly estimates of such parameters as surface parcel liftedindex values (positive area), cap strength (negative area),low-level convergence (implying localized lifting, usuallyalong a boundary; see, e.g., Wilson et al. 1988), and manyothers.

Forecasting for the longer term (e.g., the SELS 0700UTC convective outlook forecasts the potential for general

2 For example, VDUC uses 500-mb static temperatures to computesurface-based lifted index values for the first few hours after receipt ofscheduled radiosonde data, with Nested Grid Model (NGM) forecast 500-mb temperatures being used when they become available (4 to 9 h afterscheduled radiosonde data receipt).

FIG. 3. An example of a composite prognosis depicting 12-h forecastpositions of those surface and upper-air parameters that affect severethunderstorm development. Forecast is for 0000 UTC 27 April 1991based on initial data from 1200 UTC 26 April 1991. This is a black-and-white copy of a color original in Doswell et al. 1993 (their Fig. 1); thedetails of the parameter coding can be found therein and are not impor-tant here. Rather, the intent here merely is to exemplify the appearanceof a composite prognosis.

thunderstorms for a period extending out to 29 h from issuetime) primarily involves interpretation of numerical modelproducts. Typically, the SELS forecaster constructs acomposite chart of “initial” conditions as soon as data fromthe latest radiosonde release have been analyzed (seeDoswell 1982, II 31— 32). From this composite chart, thethree-dimensional relationships among the “initial”synoptic patterns and meteorological parameter fields canbe visualized.

Composite prognoses (“progs”) for 12-h intervals outto 48 h from the time of the “initial” composite chart (Fig.3) are then constructed, primarily by assimilating NWPmodel products from the NMC. Patterns and parametervalues on the resultant prog charts are utilized to prepareforecasts both for deep convection and for severe weatherevents (see sections 4, 5, and 6). Concerning deepconvective development, model parameters that relate tosuch factors as vertical motion, instability (positive area),and the capping inversion (negative area) are included onthe prog charts. Recently, a procedure was developed atNSSFC that allows the forecaster to display many of theoperational NWP fields used in constructing compositeprogs (Cope 1992). The procedure assigns a particular colorto each parameter field and has a looping capability thatallows the forecaster to see how model forecast fieldschange with time. Any combination of fields may besuperimposed (e.g., instability and vertical motion; Fig.4). Another important notion in developing the forecastsis that of limiting factors. Limiting factors allow the SELSforecaster to refine his/her perception of where deepconvection is possible. The following are some examples.

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

1. Deep convection is usually limited to those areaswhere the NGM forecast 1000—500-mb mean relativehumidity is greater than 40% to 45%.

2. Deep convection is generally limited to those areaswhere lifted index values [from combined use of theLimited-area Fine-mesh Model (LFM) and NGM output,and computed surface parcel lifted index values] are zeroor less. Since the NGM often does not forecastthermodynamic profiles accurately (particularly boundary-layer moisture), model prediction of lifted index values isfrequently “poor” (see Weiss 1987a). The SELS forecasteruses a combination of the most reliable aspects of both theNGM and LFM forecasts of lifted index values and patterns

in preparing composite prognoses. Another method ofmaking lifted index forecasts is to compute surface parcellifted index values (Hales and Doswell 1982) for specificpoints of interest. The technique requires an estimate (orobservation) of surface temperature and dewpoint at aparticular place and time, along with a forecast 500- mbtemperature (say, from the NGM) valid for the same placeand time. It should be noted that using surface-based valuesfor determining lifted index can be unrepresentative; forexample, there may be a shallow stable layer near thesurface, above which lies a deep, unstable layer. However,in spite of this limitation of a surface parcel lifted index,the result usually is more representative than the NGMforecast lifted index value.

FIG. 4. Frames from a SELS display loop that depicts forecasts of NGMbest lifted index fields (solid contours at 2°C intervals for values ≤ 0)and NGM vertical motion fields [contours at 0.1 Pa s-1 intervals, dashedcontours at 0.2 Pa s-1 for negative (upward) vertical motion] for (a) 00 h,(b) 6 h, and (c) 12 h after 1200 UTC 4 March 1992.

DECEMBER 1992 J O H N S A N D D O S W E L L

3. Deep convection is usually limited to those areaswhere the forecast 700-mb temperature (NGM) is colderthan 12°C. Deep convection also generally is limited toareas where the forecast 1000—500-mb thickness values(NGM) are 5790 m or less. These two empirical rules relateto the strength of the low-level capping inversion (between850 and 700 mb) or “negative” area. They do not apply athigh surface elevations (e.g., where 700mb becomes a partof the mixed layer), and such a strong cap still may beovercome in certain situations, such as when convectiondevelops above the cap.

4. Deep convection is usually limited to areas wherethe NGM-forecast 700-mb vertical motions are generallyneutral (> –3 to < +3 x l0-3 mb s-1) to decidedly upward(< –3 x l0-3 mb s-1; note that the sign is reversed on NGMoutput). This relates primarily to the effects of synoptic-scale vertical motion on the strength of the capping/subsidence inversion. In some cases, however, this limitingfactor may also be related to mesoscale and smaller-scalelifting mechanisms that allow a parcel to reach its LFC.NWP model forecasts for low-level boundaries andmaxima in low-level moisture convergence are morereliable indicators of such small-scale processes than themodel’s vertical motion, per se.

It should be noted that although the forecaster extractsmany of the limiting factors directly from NGM fieldswhen preparing a composite prognosis chart, the actualconvective forecast often must be adjusted to account forperceived model biases (Junker et al. 1989). Further, forthose outlooks that become valid within 12 h from the timeof issuance, a close assessment of the synoptic- andsubsynoptic-scale initial conditions and trends in surfaceand upper-air features is made prior to forecast issuance(Doswell 1 986a). This typically includes examining most“real-time” data sources (e.g., satellite imagery) andsubjectively analyzing a surface chart covering all areasof interest. Significant differences between model trendsand realtime trends are noted and the forecast is adjustedaccordingly (e.g., Hales 1979), particularly in the case ofongoing deep convection. In such situations, shallowbaroclinic boundaries often develop that may disrupt theflow of low-level moisture from the source region. Suchmesoscale developments can dramatically affect theconvective evolution over a broad region.

Recently, NMC-generated model forecast soundings(Plummer 1989) have become available. For selected citiesacross the continental United States, the forecaster candisplay model forecast thermodynamic and wind profilesat hourly intervals out to 48 h, showing how key elementsin the model atmosphere change with time (e.g., changesin the cap strength or the depth of the moist layer). Ofcourse, these soundings are only as good as the modelforecast, and have relatively coarse vertical resolution (16

levels). Since systematic biases are present in models,model sounding data may need subjective adjustment usingcomputer programs like SHARP (Hart and Korotky 1991).For example, low- level moisture can be added to themodel-forecasted boundary-layer conditions in situationswhere experience with the NWP models suggests a drybias. The forecaster then can see quantitatively how thischange would affect parameters derived from the modifiedsounding.

NSSFC recently gained the ability to modify modelforecast thermodynamic profiles using hourly surface data(Bothwell 1992). The thermodynamic profile for thesurface lifted parcel is overlaid with the model forecastprofile, so that positive and negative areas can be visualized(see Fig. 5a). This diagnostic tool has been helpful to SELSforecasters in making short-term forecast decisions.

Once the forecaster decides that deep convection islikely, the next question is whether or not this convectionwill be capable of producing large hail, damaging winds,or tornadoes. That is, the forecast task shifts fromforecasting deep, moist convection to forecasting severeconvective weather, which clearly is contingent on thepresence of deep convection.

4. Forecasting large hail

Hail development is quite complex, and the reader isreferred to Knight and Squires (1981) and Morgan andSummers (1986) for additional information. A necessaryingredient for development of large hail is a strongupdraft—that is, one that is capable of supporting theweight of a hailstone long enough for it to reach a largesize. A primary contributor to a strong updraft is thermalbuoyancy (positive area) for lifted parcels. In general, thegreater the buoyancy, the greater the potential for largehail. Over the years, this relationship has been a primaryfactor used in computations to estimate potential hailstonesize at ground level from radiosonde data (e.g., Fawbushand Miller 1953; Foster and Bates 1956).

Updraft strength, by itself, is not a sufficient indicatorthat large hail will develop. Hail development and sizeattained appear to be greatly affected by variations in storm-scale wind structures (see Nelson 1983). These variationsaffect the transit time of hail embryos through the hail-growth zone. Because of these variations in structure,supercells (or strong multicell storms) occurring in similarthermodynamic environments may differ in the size,amount, and distribution of hail produced.

Another important factor affecting hailstone size at thesurface is the effect of melting as hailstones fall throughthe freezing level to the surface. Melting is influenced bya number of subfactors, including 1) distance between thefreezing level and the ground, 2) mean temperature of the

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

DECEMBER 1992 J O H N S A N D D O S W E L L

downdraft air between the hailstone’s freezing level andthe ground, and 3) hailstone size, which affects how longit takes the hailstone to fall.

The environmental wet-bulb zero (WBZ) levelapproximates the freezing-level height of downdraft air,within which the hailstone is likely to be found. The higherthis level is, the longer the melting process can operate.Also, the higher the mean temperature between the WBZlevel and the ground, the faster the melting rate. Largehailstones fall at greater terminal velocities than smallstones; that is, for a given WBZ level, melting of smallhailstones lasts for a longer time than it does for largerstones. Moreover, melting takes place on the surface ofthe stone and surface area is proportionately greater forsmall stones than for large stones; area goes up as thesquare of the radius whereas volume goes up as the cubeof the radius. Attempts to account for melting effects areincluded in the hail-size algorithms discussed earlier.Together with the updraft component (due to buoyancy),they make up the major components of these algorithms.

Another factor influencing hail development notaccounted for by these algorithms is the nonhydrostaticpressure effects on updraft strength. Numerical simulations(e.g., Weisman and Klemp 1984; Brooks and Wilhelmson1990; McCaul 1990) strongly suggest that interaction ofthe updraft with environmental winds can createperturbation pressure gradients and resultant verticalaccelerations that contribute substantially to updraft speeds.In some instances, this contribution may be more influentialthan buoyancy in driving the updraft (Weisman and Klemp1984; McCaul 1990). Since the environmental windstructure associated with these vertical accelerations alsois associated with supercell development (see section 6),the effects would apply particularly to supercell convection.Therefore, forecasters should remember that whensupercells (or strong, well-organized multicells) arepresent, hail sizes may be considerably larger than thecurrent operational algorithms would predict.

When employed operationally, the traditional methodsof estimating hail size (e.g., Fawbush and Miller 1953;Prosser and Foster 1966) have resulted in, at best, limitedsuccess (e.g., Doswell et al. 1982). Part of the problemwith these algorithms is that they are based on soundingstaken from a widely spaced network at relatively infrequent(12 h) intervals. Therefore, their input data may not berepresentative of conditions when convective storms

develop. Leftwich (1984) has shown that, using thealgorithm of Foster and Bates (1956), the reliability of itspredicted sizes deteriorates rapidly as time advances pastsounding release time.

Recently, Moore and Pino (1990) introduced a hail- sizeforecasting algorithm that accounts for the negative effectsof water loading and entrainment on the strength of theupdraft. It is designed to be used in an interactive manner,with hail-size forecasts adjusted using advectedtemperatures aloft and the latest surface data. In preliminarytests by Moore and Pino involving cases from the plainsregion during the summer, this algorithm has shownsignificantly greater skill in forecasting hail size than thatof Fawbush and Miller (1953).

Supercells associated with relatively weak instabilityoften do not produce large hail (Johns and Sammler 1989).Operational experience suggests that such occurrences aremost common from late fall through early spring, in theregion from the lower half of the Mississippi valley to thesoutheastern United States. As an example, during theRaleigh, North Carolina, tornado outbreak of 28 November1988, no large hail [diameter 1.9 cm (3/4 inch) or greater]was reported from any of the three supercells that trackedacross portions of North Carolina and Virginia (STORMDATA 1988; and personal communication with R. Gonskifrom the NWS Forecast Office in Raleigh). Empirically, itseems that the current hail-size forecasting algorithmsperform best in “pulse” storm environments. Thecomplexity of hail development with more organizedconvection suggests that attempts to develop a forecastingalgorithm that is effective in supercell and strong multicellenvironments (see sections 5 and 6) may be a difficult task.

While forecasting for severe weather watches is focusedon hail-size estimation algorithms, for SELS convectiveoutlooks, both pattern recognition and climatology play arole. An example of a pattern typically associated withhail is the “cold low” pattern (Type D in Miller 1972).When moisture is sufficient for moderate to stronginstability in the vicinity of the cold pool aloft associatedwith an upper closed low, relatively low-toppedthunderstorms are likely during afternoon and earlyevening hours. Hail often is reported with this convectionand, if instability is sufficient, hail diameters can exceedsevere limits. Hail diameters in these situations rarelyexceed 4.4 cm (1 3/4 inch). In cold low patterns, thecommon occurrence of hail at the surface is aided by low

FIG. 5. (a) The 0000 UTC 16 March 1992 NGM forecast sounding for Oklahoma City, Oklahoma (OKC), at 2000 UTC 16 March and the actual2000 UTC surface-based lifted parcel thermodynamic profile for Fort Sill, Oklahoma (FSI) (medium line). The NGM forecast sounding shows thethermodynamic profile (heavy line), wet- bulb temperature (light line), hodograph (upper right), and a wind profile (on right; wind speeds in knots,with flag indicating direction and length of flag proportional to speed). An explanation of other parameters may be found in Bothwell (1992). (b)Magnification option for hodograph in (a); numbered squares on hodograph indicate height (AGL) in km. The first point of the hodograph representsactual 2000 UTC surface wind at FSI. The X indicates storm motion vector as estimated by multiplying the 0—6-km AGL mean wind speed by 0.75and using a direction 30° to the right of she 0—6-km AGL mean wind direction (denoted by 30R75).

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

WBZ heights and low mean temperatures below the WBZheight.

An example involving climatology can be found in latespring and summer over the central High Plains (Kelly etal. 1985). If moderate to strong lapse rates and moisture,resulting in moderate to strong instability, are forecast inan upslope flow severe thunderstorm pattern (Doswell1980), large hail is very likely to be associated with anystrong storms that develop. Therefore, in those upslopeflow cases where parameters (other than instability)supporting severe thunderstorm development are forecastto be marginal, but instability is expected to be moderateto strong, the forecaster may predict a risk of severethunderstorm development in the outlook, based solely onthe regional climatological association of those conditionswith large hail.

5. Forecasting damaging winds

Damaging straight-line winds associated with deepconvection almost always are generated by outflow thatoccurs at the base of a downdraft.3 Fujita and Byers (1977)designated exceptionally strong downdrafts as downbursts,but the term’s meaning has grown to encompass anydamaging (or potentially damaging) winds produced bydowndrafts. Fujita (1978) has noted that damagingwindstorms can occur on different scales, and has calledspecial attention to the microburst, which he defines to bea downburst ≤ 5 km in diameter.

Given that deep convection develops (see section 3),ingredients necessary for damaging winds at the surfaceare those promoting strong downdrafts. Precipitationloading and negative buoyancy due to evaporative coolingare recognized factors in initiating and sustaining adowndraft (Doswell 1982). Precipitation loading is the drageffect of liquid water, which enhances parcel descent.Essentially, the greater the quantity of liquid water per unitvolume, the greater the precipitation drag.

The other factor, negative buoyancy due to evaporativecooling, is created when precipitation falls through a layerof unsaturated air. Once a downdraft is established,continued entrainment of unsaturated air can aidevaporation. Studies indicate that most of the entrainedair originates from middle layers of the atmosphere(approximately 3—7 km AGL; e.g., see Foster 1958).Generally, downdraft strength is enhanced by availabilityand entrainment of relatively dry (low relative humidity)air.

Evaporative cooling (and downdraft strength) also isenhanced by 1) large liquid water content per unit volume,2) small drop size, and 3) a steep lapse rate, as noted by

Kamburova and Ludlam (1966) and Srivastava (1985). Thelarge liquid water content and drop-size factors relate tothe amount of liquid water surface available forevaporation. Further, a steep lapse rate acts to maintainnegative buoyancy as a downdraft parcel descends.

Downward transfer of horizontal momentum fromstrong flow aloft can enhance outflow (Brandes 1977 andothers). Generally, the stronger the environmental windsin the downdraft entrainment region, the greater thepotential contribution to outflow strength. It is unclear howmuch of this contribution actually is realized in outflowwind speeds (see Foster 1958), however.

From the viewpoint of convective wind-gustforecasting, it is important to distinguish between updraftand downdraft instability. Updraft instability invariably isassociated with positive buoyancy created by thetemperature difference between a parcel rising moistadiabatically and its environment. Thus, moisture isnecessary to ensure that ascent is along a moist adiabat,while high lapse rates create a large temperature differencebetween the parcel and the environment. The drag fromwater loading tends to diminish the effects of positivebuoyancy.

Downdraft instability, on the other hand, is primarilythe result of negative buoyancy. In downdrafts, the dragfrom water loading combines with the effects of negativebuoyancy to enhance downdrafts. Whereas virtually allunstable updrafts are saturated, unstable downdrafts mayor may not be saturated. If a parcel descends unsaturated,a high lapse rate is required; if a parcel descendsunsaturated, there must be a continuous source of liquidwater for evaporation. Otherwise, the parcel quickly willbecome warmer than its environment through adiabaticheating, thereby retarding its descent. Conditions forunstable downdrafts, therefore, do not necessarily coincidewith those for unstable updrafts. Those indices andparameters that reflect updraft instability will notnecessarily be reliable indicators of downdraft instability.

Early algorithms developed to predict outflow speeds(e.g., Fawbush and Miller 1954; Prosser and Foster 1966)attempted to account for negative buoyancy created byevaporative cooling and for momentum transfer. Theirsuccess has been limited, however, for reasons similar tothose associated with the early algorithms for forecastinghail size (see Doswell et al. 1982). A more recent algorithmby Anthes (1977) accounts for entrainment into thedowndraft as an additional component affectingevaporative cooling. Pino and Moore (1989) havedeveloped an algorithm used interactively that combinesFoster’s (1958) ideas about outflow strength with Anthes’ideas about entrainment. The Pino—Moore wind-gustforecasts are adjusted with time using advectedthermodynamic parameters aloft and the latest hourlysurface data. In preliminary testing, the algorithm shows

3 On some rare occasions, inflow winds can attain damaging propor-tions. This is usually restricted to supercells, described in section 6.

DECEMBER 1992 J O H N S A N D D O S W E L L

an increase in skill over those of Fawbush and Miller (1954)and Prosser and Foster (1966).

a. Wind gusts in weak shear environments

When the environmental shear is weak, thethermodynamic profile (pattern) is a primary signal foridentifying when strong convectively induced winds arelikely to occur. Two prototypical thermodynamic profilesassociated with strong outflow in weak environmental windsituations are 1) the “inverted V” or “Type A” profile(Beebe 1955; Barnes and Newton 1986), and 2) the weaklycapped “wet microburst” profile (Read 1987).

The inverted V profile is characterized by a deep dry-adiabatic layer from near the surface to the mid- levels, avery dry lower layer, and a moist midtropospheric layer(Fig. 6a). “Dry” microbursts (Krumm 1954; Caracena et

FIG. 6. (a) Skew T-logp plot of the 1200 UTC 15 July 1982 Denver,Colorado, upper-air sounding (after Caracena et al. 1983). (b) Skew T-logp conceptual model of sounding associated with wet downbursts innorth Texas near onset (after Read 1987).

al. 1983; Wakimoto 1985) are typically associated withvery high LFCs and only marginal instability for updrafts.Thus, convection is usually weak and pulselike in character,and electrical activity may be absent. Because such eventsusually occur in rather stagnant synoptic conditions,forecast schemes have been developed by Wakimoto(1985) and Caracena et al. (1983) based on 1200 UTCsounding analyses and expectations for diurnal heating.Forecasters have had success in employing such schemesin the wording of their afternoon and evening forecasts(see Sohl 1987) and they are considered in SELSoperations.

The wet microburst profile (Fig. 6b) typically displayshigh moisture values through a deep, surface- based layer,with the top of the moist layer sometimes extending beyond4—5 km AGL. Relative humidities above the moist layerare typically low. As diurnal heating occurs, a dry-adiabaticlayer can develop in the lower 1.5 km (5000 ft) AGL, sothere may be weak to moderate potential instability andlittle or no capping inversion. Wet microbursts are alsotypically “pulse” in nature and occur during stagnantsynoptic conditions.

A recent study by Atkins and Wakimoto (1991) suggeststhat the afternoon thermodynamic environment on activewet microburst days in a humid climate often displays aθe. (equivalent potential temperature) difference betweenthe surface and midlevels equal to or greater than 20°C(see Fig. 7). On thunderstorm days when no wet microburstactivity was observed, θe differences were less than or equalto 13°C. This suggests that the vertical θe difference wouldbe a potentially useful forecast parameter for strong windgusts. Further, initial development of thunderstorms withmicroburst potential may be anticipated by determiningwhere localized lifting is most likely, such as in the vicinityof topographic features and low-level boundaries (see Read1987).

b. Wind gusts in moderate and strong shear environments

In situations where the vertical wind shear from thesurface through midlevels is moderate or strong, convectivewind forecasting becomes more complicated. In additionto the thermodynamic profile, the wind profile and thesynoptic pattern also play a significant role in developmentof damaging winds. As vertical wind shear increases, deepconvection is increasingly likely to take the form of self-perpetuating convective systems (i.e., the outflowsystematically initiates new updrafts). These systems turnout to be responsible for most severe convective windevents. Figure 8 illustrates the range of self-perpetuatingstorm structures associated with damaging winds inmoderate to strong wind-shear situations.

On the smaller end of the scale is the isolated supercell(Fig. 8a). Damaging outflow winds usually are associated

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

with downdrafts accompanying such storms, butoccasionally damaging inflow winds occur with supercells.Typically, isolated supercells develop in a very unstableair mass exhibiting a capped “loaded gun” or “Type B”thermodynamic profile (Fig. 9) (Fawbush and Miller 1953;Barnes and Newton 1986), which is unstable for bothupdrafts and downdrafts. Winds usually veer strongly withheight, resulting in the “crossing jets” pattern shown inFig. 10a. Necessary ingredients for supercell developmentand forecasting of supercells will be discussed further insection 6.

A more common convective structure associated withdamaging wind events occurring in moderate to strongwind-shear environments is the bow echo (Figs. 8b—d),as designated by Fujita (1978). Bow echoes are typicallylarger in scale than isolated supercells, and occasionallyinclude embedded supercell circulations (e.g., Przybylinskiand DeCaire 1985; Schmidt and Cotton 1989) that mayproduce tornadoes (e.g., Smith and Partacz 1985; Molleret al. 1990) as well as strong convective gusts. Bow-shapedechoes range in scale from less than 15 km (10 mi) to over150 km (100 mi) in length, and often comprise componentsof an extensive convective line (see Fig. 8d). Some larger-

FIG. 7. Schematic θe profiles on days when the environment is conductive for wet microburst occurrencein a humid region (after Atkins and Wakimoto 1991).

FIG. 8. (a) Schematic representation of flow associated with supercell thunderstorm (after Lemon and Doswell 1979), showing forward-flankdowndraft (FFD), and rear-flank downdraft (RFD). (b) Schematic representation of downdraft flow associated with a relatively large bow echo. (c)As in (b) except for line-echo wave pattern form. Eastword extension from north end of bow echo is known as a warm advection wing (Smith 1990).(d) as in (b) except for extensive squall line with embedded bow echoes and line echo wave patterns.

scale bow echoes have smaller-scale bow echoes embeddedwithin them. The curved bow-echo structure (Fig. 8b)apparently reflects the diverging outflow winds associatedwith a strong downdraft. Occasionally, a persistent large-scale bow echo or series of bow echoes produces asuccession of downbursts that affect a widespread area (orswath). This larger-scale wind event has been called afamily of downburst clusters (Fujita and Wakimoto 1981),or more recently, a derecho (Johns and Hirt 1987).

There are two basic synoptic patterns associated withbow-echo development: I) the warm-season pattern, and2) the dynamic pattern (Johns 1993). The warm- seasonpattern is most commonly observed during late spring andsummer and is usually associated with progressivederechos (Fig. 11 a). The convective activity develops inan area of low-level warm advection, along or north of aquasi-stationary boundary (Maddox and Doswell 1982;Johns 1984). An elevated rear-inflow jet (see, e.g.,Augustine and Zipser 1987; Smull and Houze 1987) isoften found with these events, and it appears to be a criticalfactor in the initiation and maintenance of warm-seasonderechos (e.g., Burgess and Smull 1990; Schmidt et al.1990).

DECEMBER 1992 J O H N S A N D D O S W E L L

Once formed, bow echoes move along the boundary,but usually slightly to the right of the mean flow. The airmass along the boundary in warm-season synoptic patterncases is typically extremely unstable (Johns et al. 1990a),owing to high lapse rates in elevated mixed layers (seeLanicci and Warner 1991) and “pooling” of low-levelmoisture (i.e., development of an area or zone wheredewpoints are higher than in the surrounding region) inthe convergence zone near the boundary (Fig. 12). Thisstrong instability (see, e.g., Fig. 9) appears to play a rolein maintaining the elevated rear- inflow jet and the strongoutflows near the surface gust front (Weisman 1990). Thecap associated with the dry, high lapse-rate air in the lowermidtroposphere plays a significant role in restrictingsouthward development of bow-echo activity in the warmsector. Therefore, bow-echo structures in this type ofpattern usually do not develop into extensive lines. Instead,they tend to take relatively short bow-echo or line-echowave pattern (or LEWP; see Nolen 1959) forms (see Figs.8b and 8c).

During the past few warm seasons, SELS forecastershave mentioned occasionally the possibility of progressivederecho development in the convective outlook when theircomposite prognoses indicate patterns and parametervalues that favor such development. About once or twicea year, parameters known to favor warm-seasonprogressive derecho development are quite strong andaffect an extensive area. When a bow-echo complex

FIG. 9. Composite skew T-logp plot of sounding (after Fawbush andMiller 1953) associated with tornadoes in t he plains region, often referredto as the “loaded gun” or “Type B” sounding.

develops in such a situation, SELS forecasters have theoption of inserting “enhanced wording” into watches issuedahead of the bow-echo complex. Figure 13 shows anexample of an enhanced wording watch issued during aparticularly intense progressive derecho event.

The other basic synoptic pattern associated with bow-echo development is the dynamic pattern (Johns 1993).This pattern is usually associated with an extensive squallline that results in a serial derecho (Fig. 11 b). It may occurat any time of year, but appears to be least common duringmid- and late summer. It has some aspects of the classicsevere weather outbreak (Forbes et al. 1980), usuallyinvolving a strong, progressive low pressure system (Fig.10a). Since many cases occur during the cool season,however, the warm- sector air mass may be only marginallyunstable and the vertical wind profile typically exhibits a

FIG. 10. (a) Idealized sketch (after Barnes and Newton 1986) of amiddle-latitude synoptic-scale situation favorable for development ofsevere thunderstorms. Solid thin lines denote isobars, broad arrowsrepresent low-level jet (LJ), polar jet (PJ), and subtropical jet (SJ). (b) Asin (a), except for the situation associated with long-lived warm-seasonprogressive derechos. The line denoted by B–M–E is the derecho’s longaxis, with B the beginning point (after Johns et al. 1990b).

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

pattern where upper- and low-level jets are moreunidirectional than in the classic outbreak pattern (Johns1993).

c. Climatology

Dry microbursts and inverted V soundings are mostfrequent on the High Plains and interior portions of thewest, while wet microbursts are most common east of the

FIG. 11. (a) Schematic representation of features associated with a progressive derecho near midpoint of its lifetime. The total area affected byderecho during its lifetime is indicated by hatching; frontal and squall-line symbols are conventional. (b) As in (a), except for features associated witha serial derecho (after Johns and Hirt 1987).

FIG. 12. Composite 850-mb pattern near the initiation time of long-lived warm-season progressive derechos (after Johns et al. 1990b). Thesolid arrow B–M–E is the derecho’s long axis, with B its beginning point;the broad dashed line is a trough line (or 850-mb front); isotherms (dashed)and selected idodrosotherms (solid) are in °C. Full wind barbs are equalto 10 kt (5.1 m s-1) and half-barbs equal 5 kt (2.6 m s-1).

FIG. 13. Example of a severe thunderstorm watch that contains“enhanced wording” for expected intense derecho winds. The watchwas issued by the National Severe Storms Forecast Center during thelong-lived tornado event of 7–8 July 1991.

High Plains. Both types of pulse events are most commonin summer and typically occur during the afternoon andearly evening.

Isolated supercell and bow-echo wind events are mostfrequent in the afternoon and evening, but are less diurnallydependent than pulse storm microbursts. Isolated supercellsare most likely in the plains during spring, while warm-season bow-echo events are most likely in late spring and

DECEMBER 1992 J O H N S A N D D O S W E L L

summer in a band from the northern plains to the Ohiovalley (Fig. 14). Operational experience suggests thatdynamic synoptic pattern bow-echo events occurringduring fall, winter, and early spring are most likely fromthe lower and mid-Mississippi valley eastward to the EastCoast and are less diurnally dependent than other windevents.

6. Forecasting tornadoes

Tornadoes can be subdivided into two basic groups:those associated with supercells (e.g., Fig. 8a) and thosethat are not (e.g., as defined by Wakimoto and Wilson1989).4 Nonsupercell tornadoes are just beginning to beunderstood, and while some forecasting methods have beendeveloped for them, such methods at present are tied to

FIG. 14. Total number of derechos occurring in 2° latitude by 2° longitude squares during May through Augustfor the period 1980––1983 (after Johns and Hirt 1987).

topographic features in specific locations. Supercelltornadoes include most of the strong and violent tornadoevents (F2 through FS, as classified by the F-scale systemof Fujita 1971) and account for a disproportionate shareof all tornado-related deaths, injuries, and damage. Owingto recent storm observations and numerical modelingexperiments, much has been learned about supercellstorms. The remainder of this section will primarily addressthe nature of supercell storm environments and currentmethods of forecasting supercell-induced tornadodevelopment.

a. Storm structures associated with supercell circulations5

It appears that deep, persistent mesocyclones that seemto be associated with supercell tornadoes can occur within

4 As with any classification scheme, we not that not every storm fitsneatly into a category. See Doswell (1991b) for a discussion of classifi-cation schemes and their potential pitfalls.

5 Supercell circulation is used here as a general term to include allconvective storms with deep, persistent mesocyclones, regardless of theecho morphology depicted by radar reflectivity data.

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

FIG. 15. Examples of complex convective structures (as indicated by radar reflectivity imagery) associated with strong and violent tornadoes: (a)F4 intensity McColl, South Carolina, tornado of 28 March 1984 (indicated by ‘6”) associated with small bow echo (after Storm Data for March1984); (b) F4 intensity Allendale, Illinois, tornado of 7 January 1989 (indicated by “T”) associated with a short line (after Przybylinski et al. 1990);(c) F3 intensity Maninsville, Indiana, tornado of 10 March 1986 (indicated by “T”) associated with bow echo and possible line-echo wave pattern(after Przybylinski 1988) and (d) F4 intensity Chesnee, South Carolina, tornado of 5 May 1989 (indicated by white “+” for 2220 UTC) associatedwith a spiral band. Additional F4 tornadoes occurred at 2254 and 0001 UTC as indicated (after July 1990).

a broad range of storm structures (Doswell et al. 1990;Doswell and Burgess 1993). Further, recent observationalevidence suggests that “classic” isolated supercells (Fig.8a) account for less than half of all strong and violenttornadoes occurring in the United States (Johns et al. 1993).It appears that a majorityof such events are associated with a variety of complexstorm structures (as seen on radar), including bow echoes,LEWPs, and spiral-banded echoes (e.g., Fig. 15), virtuallyall of which contain a deep, persistent mesocyclone, butwhich do not fit traditional models of supercells. The high-precipitation supercell category, described in Moller et al.(1990), has been developed, in part, to account for thesecomplex structures within a spectrum of supercell storms(Doswell et al. 1990).

b. Wind environments associated with supercelldevelopment

Observations and, especially, numerical modelsimulations suggest that the most critical environmentalfactors affecting supercell development involve thestrength and nature of tropospheric winds, particularly inthe lower and middle layers. Numerical modelingexperiments suggest that 1) the nature of the wind profilein the storm inflow layer (Weisman and Klemp 1984), 2)the strength of storm-relative inflow (Lazarus andDroegemeier 1990), and 3) the strength of wind shearthrough midlevels (Brooks et al. 1993) are all importantfor development and maintenance of supercell storms.

DECEMBER 1992 J O H N S A N D D O S W E L L

1) ROTATIONAL POTENTIAL OF THE STORM INFLOW LAYER

In an operational setting, it is difficult to define preciselywhat constitutes the storm inflow layer. Differing layers,generally within 4 km AGL, have been used to estimatethe source of air entrained into the updraft (Browning andLandry 1963).6 A study by Bluestein et al. (1989) suggeststhat, at least in very unstable “loaded gun” thermodynamicenvironments, entrainment into the updraft core is minimalabove the LFC. Therefore, it is possible that in many casesthe depth of the inflow layer is roughly coincident withLFC height (Davies and Johns 1993). Bluestein and Parks(1983) found that during spring, the average LFC heightin supercell situations in the southern plains is slightlygreater than 2 km AGL. In more moist and weakly cappedsituations (e.g., tropical cyclone environments) the LFCheight may be significantly below 2 km AGL (e.g., McCaul1991).

Streamwise vorticity (i.e., that part of the horizontalvorticity vector parallel to the flow) in the storm inflowlayer results in updraft rotation when this air is ingestedinto an updraft (Davies-Jones 1984). The characteristicenvironment rich in streamwise vorticity has a hodographthat is strongly curved, with rapidly increasing speeds inthe lower 2 or 3 km AGL (Fig. 16). Typically, hodographsassociated with supercell development in the NorthernHemisphere (Southern Hemisphere) curve to the right(left). This requires veering of the shear vector, whichproduces the curved hodograph (see Doswell 1991 a).Sometimes, supercells occur with relatively straight (onlyslightly curved) low-level hodographs when wind shear issufficiently strong through a deep layer (e.g., Charba andSasaki 1971). Therefore, the shape of the 0—3-km AGLsegment of environmental hodographs associated withmesocyclone-induced tornadoes in the United States variesfrom only slightly to very strongly curved to the right (e.g.,see Fig. 17).

FIG. 16. Typical wind hodograph associated with supercellenvironments during the alberta hail studies project (after Chisholm andRenick 1972).

6 In cases where convection is occurring above a frontal inversion, theinflow layer may have a base that is above the surface.

Two parameters related to the inflow layer rotationalpotential are 1) positive mean shear (Davies. 1989; Daviesand Johns 1993) and 2) storm-relative helicity (Lilly 1986;Davies-Jones et al. 1990). Positive mean shear estimatesthe mean shear associated’ only with straight-line or right-turning segments of hodographs in the lowest 2 km AGL(approximating the inflow layer). Since this parameterevaluates the hodograph in a ground-relative framework,its applicability is affected by observed storm motions.Figure 18 illustrates the range of 0–2-km AGL positiveshear values associated with a large dataset of 242 strongand violent tornadoes (occurring between April 1980 andMarch 1990) that was assimilated by Johns et al. (1990b,hereafter referred to as JDL).

Helicity, which is considered in a storm-relativeframework, is a computationally more stable parameterthan positive mean shear (Davies-Jones et al. 1990).Helicity can be visualized as twice the area swept out bythe storm-relative motion vector in a layer (see Davies-Jones et al. 1990; Doswell 1991a). Therefore, the stormmotion vector plays an important role in evaluating thisparameter. Helicity values using observed storm motionshave been calculated for limited data samples (Davies-Jones et al. 1990; Davies and Johns 1993), with resultsindicating that most strong and violent tornadoes in thesedatasets were associated with values in the 0–3-km layergreater than 300.

FIG. 17. Ground-relative hodographs (prepared by Jon Davies of Pratt,Kansas) for low-level wind structure (0–3 km AGL) at (a) Nashville,Tennessee, at 1200 UTC on 24 December 1988, and for low and midlevelwind structure (0–6 km AGL) at (b) Omaha, Nebraska, at 0000 UTC on29 July 1986. The midlevel portion of the hodograph for Nashville wasnot available. Each ring increment represents 10 kt (5.1 m s-1).

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

FIG. 18. Scatter diagram showing combinations of CAPE in J kg-1

and 0–2 km AGL positive wind shear (x10-3 s-1) for all 242 tornadoes inthe JDL datasheet. Solid curved line is a suggested lower limit ofcombined CAPE / low-level shear values that would support developmentof strong and violent tornadoes (after Johns et al. 1993).

Forecasting storm motion can be a formidable task,since storm motions vary considerably from case to case(Davies and Johns 1993). Attempts have been made toassume an average storm motion based on the mean windthrough deep layers of the troposphere [e.g., the cloud-bearing layer used by Maddox (1976) and the 0—6-kmAGL layer used by Davies and Johns (1992)]. These “meanwind” methods for forecasting storm motion have met withlimited success, at least in part because storm motions canbe affected by other factors (e.g., Maddox et al. 1980; Johnset al. 1993) than the environmental flow.

Operational experience suggests that helicity (and low-level positive shear) values exhibit a diurnal oscillation,with a tendency for 1200 UTC values to be higher thanthose at 0000 UTC. This is particularly noticeable whendynamic forcing is weak, and appears to be related to thediurnal changes in the strength and nature of the low-leveljet (e.g., Bonner 1968; Frisch et al. 1992). This diurnallow-level wind structure change directly affects helicity(Maddox 1993), which must be considered whenforecasting updraft rotational potential.

Several forecasting methods involving parameters usingwind forecasts by the operational NWP models have beendeveloped or proposed (e.g., Woodall 1990; Davies andJohns 1993). Typically, these methods involve calculatinghelicity values from forecast winds for a variety of potentialstorm motions. A recently developed program at SELS

displays hourly forecast hodographs and computes helicityvalues by using an assumed storm motion vector, the latesthourly surface wind, and NGM wind forecasts for selectedpoints (see Fig. 5b). An experiment by Piltz (1992) suggeststhat hourly areal maps of helicity based on forecast (orobserved) motions, surface data, and the lowest layers ofwind profiler data (Dunn 1986) may become a helpfulshort-term forecast tool. These techniques have had limitedevaluation, but appear to be most successful when theenvironmental winds are moderate to strong.

2) STRENGTH OF THE STORM INFLOW

A recent modeling study by Lazarus and Droegemeier(1990) suggests that storm-relative low-level inflowstrength is a critical factor in the development of a stronglyrotating updraft. Inflow strength is related to both stormmotion and the strength and direction of winds in the inflowlayer. One can estimate the potential environmental windcontribution to storm inflow by subtracting the meantropospheric wind vector (which often is related to stormmotion) from the mean wind vector in the storm inflowlayer. The greater the resulting value is, the greater therange of storm motions that would result in sufficientinflow for strong rotation.

One also can determine the potential strength of storm-relative inflow in a particular situation by computing thedifference between an estimated (e.g., Maddox 1976) orobserved storm motion vector and the mean wind vectorfor the inflow layer. The SHARP program (discussed insection 3) calculates inflow vectors and storm-relative layermean inflow values.

3) STRENGTH OF SHEAR THROUGH MIDTROPOSPHERE

Sufficient wind shear through midlevels can beimportant to supercell development for two reasons: 1) itremoves precipitation from updrafts by enhancing storm-relative flow, and 2) interaction of this deeper shear withstorm updrafts can induce vertical perturbation pressuregradients (Rotunno and Klemp 1982) that increase updraftstrength. Further, wind strength (and shear) in the midlevelsaffects storm motion, thus indirectly affecting storm inflowstrength (Brooks et al. 1993). Most 0–6-km AGL (84%)and almost all midlevel (3-6-km AGL) mean wind speeds(96%) in the JDL dataset are greater than 15 m s-1 (30 kt),as noted by Davies and Johns (1993).

Most of the time when wind speeds in the midlevelsappear to be too weak to support supercell development,0—2-km AGL positive shear and helicity values are alsoweak. On some occasions, however, 1200 UTC soundingsmay indicate the presence of a low-level jet in an area wheremidlevel winds are weak. The accompanying hodographsdisplay a “spike” signature (see Fig. 19) that can alert the

DECEMBER 1992 J O H N S A N D D O S W E L L

forecaster that, while the low-level shear and helicityparameters are relatively strong, the wind environment maynot be conducive to development of strong rotation. Incontrast, note that even though the hodograph associatedwith the violent tornado of 28 July 1986 (Fig. 17b) resultsin a relatively low value of observed storm-relative helicity(and positive shear), the midlevel segment (3–6 kin)displays increasing wind speeds with height.

c. Instability associated with supercell development

The JDL dataset (Fig. 18) suggests that supercells occurin environments with an extremely wide range of CAPE,varying from 200 to 5300 J kg-1. An explanation for thedistribution of CAPE values in Fig. 18 is related to therespective seasonal availability of both wind and instabilityenvironments favorable for mesocyclone development.During the cool months, wind environments favorable formesocyclone development are common and widespread.However, instability sufficient for thunderstormdevelopment is infrequent, and when it is present, valuesoften are relatively low. The reverse is true for the warmestmonths, when moderate to high values of instability areoften widespread, but wind environments supportive ofmesocyclone development are infrequent. It is notsurprising, then, that most cases in the JDL dataset exhibitintermediate values of both instability and favorable windenvironment parameters (see Figs. 18 and 20, and Fig. 2in Davies and Johns 1993), and occur in the transitionseason of spring, when the areal patterns of instability andfavorable wind environments most frequently coincide.

Figures 18 and 20 also suggest that many supercellsoccur in both very low and very high instabilityenvironments, however, with high shear associated withthe low instability cases and vice versa. This tendency hasbeen recognized by Rasmussen and Wilhelmson (1983)and Turcotte and Vigneux (1987), among others, leadingto a proposed association between storm type and the bulkRichardson number (BRN; see Weisman and Klemp 1984).

FIG. 19. Schematic representation of a ground-relative 0–6-km AGL“spiked” hodograph (see text for definition) in which midlevel flow isusually insufficient for supercell development. Each ring incrementrepresents 10 kt (5.1 m s-1) (prepared by Jon Davies of Pratt, Kansas).

In low instability/strong shear cases (BRN < 15, belowthe supercell thresholds proposed by Weisman and Kiemp),mesocyclones usually are associated with complexconvective structures (see Fig. 15), as noted by Johns etal. (1993). Some of the high instability/weak shear cases(BRN > 45, exceeding the supercell threshold proposedby Weisman and Klemp) are associated with bow echoesand isolated supercells that either are moving faster thanthe 0—6-km AGL mean wind (Smith 1990; Korotky et al.1993), or are extreme right-moving super- cells (Darkowand McCann 1977). In the former case, the environmentallow-level flow is usually relatively light and, in the lattercase, storms often move more directly into the low-levelflow. In either case, storm- relative helicity and inflow oftenare enhanced by deviant cell movement.

d. Supercells and tornadoes

Forecasting supercells is not equivalent to forecastingtornadoes. For example, Doppler radar observations at theNational Severe Storms Laboratory suggest that only about50% (or perhaps less) of all Doppler-detectedmesocyclones occurring in the southern plains in springproduce tornadoes (Burgess and Lemon 1990). One factorimportant for tornado generation is the strength of theoutflow from storm downdrafts (see Fig. 8). Evaporativelycooled downdraft outflow results in a baroclinicallygenerated, low-level contribution to vorticity that is criticalfor tornadogenesis (Rotunno and Klemp 1985; Davies-Jones and Brooks 1993). Some recent modeling results(Brooks et al. 1993) also suggest that if the outflow is toostrong, it terminates tornadogenesis prematurely. Thedetails of supercell tornadogenesis are not entirely resolvedas yet, so the development of refined forecasting techniquesmust await further research. Generally, SELS forecastersconsider that the presence of at least some “dry”(nonsaturated) air in the downdraft entrainment region isnecessary for both bow echo—induced damaging winds(see subsection Sb) and supercell tornado development.Further, from the SELS perspective, the likelihood ofsupercell development (see subsections 6b and c) is themajor factor in choosing between severe thunderstorm andtornado watches, except in cases (described next) wherenonsupercell tornadoes can be anticipated reliably.

e. Pattern recognition

Pattern recognition continues to play an important rolein tornado forecasting, and is particularly important foridentifying potential outbreaks. The “classic” tornadooutbreak exhibits a synoptic pattern similar to Miller’s(1972) Type B (Fig. 10a) tornado pattern, and ischaracterized by an unusually strong, progressive

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

FIG. 20. Scatter diagram (after Johns et al. 1993) showing combinations of CAPE (J kg-1) and 0–2-km AGL helicity (m2 s-2) utilizing the 20R85/30R75 storm motion assumptions for 242 strong and violent tornadoes of JDL dataset. All triangles (open and solid) represent cases in which theassumed storm motion is 30R75, while the assumed storm motion for the remainder of the cases is 20R85. The open circles and open trianglesrepresent violent tornadoes (F4-F5). The crosses represent cases associated with tropical cyclones.

DECEMBER 1992 J O H N S A N D D O S W E L L

extratropical cyclone. Typically, in such situations, anupper-level jet streak is associated with corresponding windmaxima at mid- and low levels, and a vertical wind profilefavorable for supercell development results, as the mid-and low-level jet maxima come into proximity (Uccelliniand Johnson 1979; Uccellini 1990). Tornado potential withthis pattern is enhanced if the associated upper shortwavetrough is moving rapidly, if it is negatively tilted, and/or ifthere is significant upper diffluence ahead of the trough.These empirical factors likely relate to developing ajuxtaposition of synoptic-scale upward vertical motion anda steepening lapse rate (McNulty 1978), rapid influx oflow-level moisture, and a supercell-favorable wind profile(i.e., winds are strong at all levels and veer strongly withheight). When all of these factors are present, the result iswhat Doswell et al. (1993) call a “synoptically evident” or“big” tornado day (in the United States there are about 10such days per year). When the SELS forecaster’s compositeprognoses indicate that all ingredients for a “big” day arelikely to become coincident, the potential for a majortornado outbreak is highlighted in the SELS convectiveoutlooks and a “public severe weather outlook” stressingthe threat is issued. If short-term indicators continue to befavorable as the day in question progresses, enhancedwording that emphasizes the outbreak potential is includedin tornado watches for the affected area (see Fig. 5 inDoswell et al. 1993 for an example).

Most other tornado episodes tend to be more localizedthan the “big” days, but still may include strong or violenttornadoes. These can occur with a variety of synopticpatterns. Many exhibit some variation of the “synopticallyevident” outbreak patterns. Typically, in such cases, either1) the necessary parameters are in juxtaposition but one(or more) of the parameters is marginal for tornadodevelopment (e.g., the wind maximum in the low-level jetmay be only marginally strong), or 2) all of the necessaryparameters are strong, but are not in juxtaposition (e.g.,the maximum in the low-level jet may not be in closeproximity to the maximum in the midlevel jet).

Also, some synoptic patterns associated with localizedtornadic episodes differ considerably from the“synoptically evident” patterns. Two such atypical patternsare Miller’s (1972) “warm advection pattern” (Type C)and “cold low/occluded front pattern” (Type D). In thewarm advection pattern, tornado activity occurs near aquasi-stationary or warm frontal boundary oriented roughlyparallel to the midlevel flow. The associated shortwavetrough is usually weak to moderate in intensity. Operationalexperience suggests that from late winter into early spring,this pattern is most often associated with a westerlysubtropical jet stream across the Gulf Coast states (e.g.,Branick 1981). From late spring into summer, this patternis commonly associated with “northwest flow” outbreaks(e.g., Johns and Leftwich 1988; LaPenta et al. 1990).

The cold low / occluded front tornado pattern (Type D)is associated with a cold upper low and / or an occludedfront that is in proximity to a cold upper low. Givensufficient instability, tornadoes may occur withthunderstorms developing near the upper low (Cooley1978; Goetsch 1988). These events develop from relativelylow-topped convection that is usually nonsupercellular innature. Tornadoes also may occur along the occluded fronttoward the warm sector (e.g., Carr and Millard 1985;Moore and Elkins 1985). Typically, the wind profilebecomes more favorable for supercells away from theupper low and toward the midlevel jet.

A localized, topography-dependent tornado pattern isthe “Los Angeles basin pattern” (Hales 1985). When adeep, occluded low develops off the central/southernCalifornia coast, the coastline shape and inland terrainpromote development of a favorable low-level wind profilefor supercell development in the Los Angeles basin. Thereare likely other areas in the United States where specificsynoptic patterns and local topography combine to inducefavorable wind profiles for supercells (e.g., see Braun andMonteverdi 1991).

Nonsupercell tornadoes also may be favored by uniqueterrain and wind patterns. For example, Szoke andAugustine (1990) have described the “Denver cyclone”pattern, which often includes tornado events. The Denvercyclone is a mesoscale surface low pressure center (orzone) and an associated low-level convergence boundaryzone that develops near Denver, Colorado, prior toconvection in certain synoptic-flow regimes. Suchphenomena may be found in other places as well, but ifso, they are not documented.

Thermodynamic and wind profiles can display commonpatterns for a particular type of tornado situation. Forexample, tornado outbreaks occurring in the central andsouthern Great Plains in spring are typically associatedwith a “loaded gun” thermodynamic profile (Fig. 9) and ahodograph that curves markedly to the right in the first 2or 3 km AGL, while being generally straight withincreasing wind speeds above 3 km AGL (e.g., Fig. 16).On the other hand, thermodynamic and wind profilepatterns associated with tropical cyclone tornadoes (Fig.21) are strikingly different from those associated with bothGreat Plains springtime tornado outbreaks and Oklahomasupercells (McCaul 1991). The composite hodograph intropical cyclone cases displays a very large loop thatapproaches being a circle. This wind profile appears to behighly supportive of supercell development but thecomposite thermodynamic profile displays relatively moistair through midlevels and has relatively little CAPE. Insuch situations, primary forecast problems often arewhether instability will be sufficient for deep convectionto develop and whether there will be sufficient dry air inthe downdraft entrainment layer to support development

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

FIG. 21. Skew T-logp plots and hodograph diagrams (after McCaul1991) for the tropical cyclone close-proximity composite (heavy line)and Oklahoma supercell composite of Bluestein and Jain (1985) (lightline). The u and v components of the Oklahoma composite are relativeto true zonal and meridional directions; boxes indicate 0–6-km AGL meanwinds.

FIG. 22. Diurnal and seasonal frequency for all 275 strong and violentFlorida tornadoes occurring from 1955 through 1983 (from the NSSFCdatabase). Numbers in the grid represent cases occurring during aparticular 3-h period of the day for each month.

of downdrafts and associated low-level baroclinic vorticityfor tornadogenesis.

f. Climatology

Several aspects of tornado climatology can alert theforecaster to those times and areas where he / she musthave increased awareness of the potential tornado threat.For example, tornado outbreaks (10 or more events)occurring in winter are most likely from the Gulf coastalstates into the mid—Mississippi valley (Galway 1977;Galway and Pearson 1981). In Florida during the winterand early spring, most tornadoes of strong or violentintensity occur between midnight and noon LST (Fig. 22).Further, Maddox (1993) has shown that most F3 or greaterintensity United States tornadoes that are reported between

0600 and 1200 UTC (0000 and 0600 CST) occur in theGulf Coast states. Tornadoes occurring in the westernUnited States coastal regions are most common from fallthrough early spring (McNulty 1981). Tornadoes associatedwith tropical cyclones affecting the United States are mostlikely in late summer and early fall and most typically occurwith those cyclones that move inland from the Gulf ofMexico or that portion of the Atlantic coastline south of320 latitude (Weiss 1987b). It is accepted generally thatNorthern Hemisphere tornadoes occur most typically withsouthwesterly flow aloft in the warm sector (see Fig. 10a);however, synoptic climatologies indicate that northwesterlyflow in the warm sector often is associated with tornadoesthat occur in late spring and summer from the upperMississippi valley to the Mid-Atlantic states (Johns 1982;Giordano and Fritsch 1991).

These are some examples of where regional tornadoclimatology indicates a significant variation from theoverall climatology (e.g., Kelly et al. 1978). Thisinformation allows a forecaster in a particular region toadjust his/her climatological expectations accordingly.

7. Discussion

Knowledge of the processes and parameters associatedwith severe local storms has advanced rapidly in recentyears. These advances have resulted primarily from acombination of numerical model simulation experimentsand observational studies. This increased understandinghas led to several new forecast techniques that are aidingSELS forecasters in assessing the potential for severe localstorm development. This is particularly true for supercelltornadoes, “pulse” storm microbursts, and bow-echodamaging winds.

New observational technology is aiding SELSforecasters in the “diagnosis and trend” aspect of assessing

DECEMBER 1992 J O H N S A N D D O S W E L L

severe local storm potential. New datasets include cloud-to-ground lightning strikes, aircraft-measured winds, WSR-88D radar imagery, wind profilers, and soon. An increaseddensity (in populous areas) of surface data from automatedstations is currently available for some areas of the UnitedStates, and these new data sources are expected to becomenational in scope within the next few years. Regionalmesoscale networks of surface stations (e.g., Crawford etal. 1992) also may become available in the near future.All these new data sources are helping remove broadtemporal and spatial gaps in the operational data network.Because subsynoptic-scale analysis has been hampered bylack of adequate operational data, SELS forecasters havefound immediate utility in these new data sources eventhough national coverage is incomplete and some sourcesare “experimental” (e.g., see Fig. 1 a).

Another development in the last few years that has aidedin the assessment of severe local storm potential is accessto the full range of numerical model outputs, allowingdevelopment of model forecast fields that are relevant tothe severe local storm forecast problem (e.g., soundingand hodograph forecasts). Further, development of morepowerful computers (and software) is allowing morecomprehensive diagnosis of the forecast data, therebymaking the output more useful.

The new forecast techniques require that the forecasterexamine existing and model forecast conditions in moredetail, while the new observational data sources and modelproducts provide massive amounts of data for the forecasterto sort through. Given the time constraints of forecasting,operational meteorologists are increasingly feeling as ifthey are “drowning in a sea of data.” The keys are to focuson those data that are relevant to the situation, and to displaythem in a useful manner. Development of procedures todo this is an ongoing process that continues as newinterpretation techniques and data sources come intooperational practice.

Our changing perceptions of the complex processesassociated with development of severe local storms makethe utility of fixed meteorological checklists (e.g.,Colquhoun 1987) problematic. However, some form of adata-management checklist or priority list may become anecessity if the forecaster is to effectively use all the newinterpretation techniques and data sources to arrive at moreaccurate and timely forecasts.

If the data management and interpretation challengescan be met, some continued improvement of the parameterassessment and pattern recognition aspects of forecastingis likely. Such improvements, however, will depend onthe ability to achieve an accurate current analysis ofmeteorological conditions and the ability of operationalNWP models to forecast changes in critical severe weatherparameters accurately. One can be hopeful that newobservational technologies that are being implemented

operationally (e.g., WSR-88D Doppler radars), areundergoing experimental evaluation (wind profilers), orare likely to be evaluated in the near future (thermal andmoisture profilers) will help to fill the temporal and spatialdata gaps. Also, experiments by numerical modelers shouldlead to operational synoptic- and mesoscale modelforecasts that have greater accuracy in predicting severeweather parameters. The pattern recognition andclimatological aspects of severe weather forecasting likelywill continue to be important to the forecasting process inthe future, however, since they help alert the forecaster towhen and where he / she should concentrate more closelyon the potential for severe weather development.

Acknowledgments. The authors especially appreciatethe numerous beneficial suggestions offered by Steve Weissof NSSFC. Fred Ostby, Jim Henderson, John Hart, andBill Sammler (all of NSSFC), Josh Korotky (OSF TrainingBranch), Harold Brooks (NSSL), Steve Lyons (NWSSouthern Region Headquarters), Jon Davies, and theanonymous reviewers also made many constructivesuggestions. The authors are grateful for the assistance ofPhillip Bothwell, Alan Cope, Jon Davies, and RonPrzybylinski in preparing the figures, as well as to DeborahHaynes and Kevin McCarthy for helping with manuscriptpreparation.

REFERENCES

Anthes. R. A., 1977: A cumulus parameterization schemeutilizing a one-dimensional cloud model. Mon. Wea.Rev., 105, 270—286.

Anthony, R. W., 1988: Tornado/severe thunderstormclimatology for the southeastern United States.Preprints, 15th Conf Severe Local Storms, Baltimore,Amer. Meteor. Soc., 511—516.

Atkins, N. T., and R. M. Wakimoto, 1991: Wet microburstactivity over the southeastern United States:Implications for forecasting. Wea. Forecasting, 6,470—482.

Augustine. J. A.. and E. J. Zipser. 1987: The use of windprofilers in a mesoscale experiment. Bull. Amer.Meteor. Soc., 68, 4—17.

Barnes, S. L., and C. W. Newton, 1986: Thunderstorms inthe synoptic setting. Thunderstorms: A Social,Scientific, and Technological Documentary. Vol. 2:Thunderstorm Morphology and Dynamics, 2ndedition, E. Kessler, Ed., University of Oklahoma Press,75—111.

Beebe. R. G., 1955: Types of airmasses in which tornadoesoccur. Bull. Amer. Meteor. Soc., 36, 349—350.

Bluestein. H. B., and C. R. Parks, 1983: A synoptic andphotographic climatology of low-precipitation severe

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

thunderstorms in the Southern Plains. Mon. Wea. Rev.,111, 2034—2046.

____ and M. H. lain, 1985: Formation of mesoscale linesof preGipitation: Severe squall lines in Oklahomaduring the spring. J. Atmos. Sci., 42, 1711—1732.

____ E. W. McCaul, Jr., 0. P. Byrd, and 0. R. Woodall,1989: Mobile sounding observations of a tornadicthunderstorm near the dryline: The Gruver, Texas,storm complex of 25 May 1987. Mon. Wea. Rev., 117,244—250.

Bonner, W. D., 1968: Climatology of the low-level jet.Mon. Wea. Rev., 96, 833—850.

Bothwell. P. D., 1992: An interactive skew-Tprogram forcombining observed and model data. Preprints, 8thInt. Conf. on Interactive Information and ProcessingSystems for Meteor.. Oceanogr., and Hydrol., Atlanta,Amer. Meteor. Soc., 135—140.

Brandes, E. A., 1977: Gust front evolution and tornadogenesis as viewed by Doppler radar. J. Appl. Meteor.,16, 333—338.

Branick, M. L., 1981: Meteorological analysis of a severesupercell thunderstorm over southern Mississippi, 3 IMarch 198 I. Southern Topics, July 1981, SouthernRegion Headquarters NWS, Ft. Worth, Texas, 7 pp.

Braun, S. A., and J. P. Monteverdi, 1991: An analysis of amesocyclone-induced tornado occurrence in northernCalifornia. Wea. Forecasting, 6, 13—31.

Brooks, H. E., and R. B. Wilhelmson, 1990: The effects oflow-level hodograph curvature on supercell structure.Preprints, 16th Conf Severe Local Storms, KananaskisPark, Alberta, Canada, Amer. Meteor. Soc., 34—39.

____, C. A. Doswell III, and R. P. Davies-Jones, 1993:Environmental helicity and the maintenance andevolution of low-level mesocyclones. Proc., TornadoSymp. III, C. Church, Ed., Amer. Geophys. Union, (inpress).

Browning, K. A., and C. R. Landry, 1963: Airflow withina tornadic storm. Preprints, 10th Wea. Radar Conf,Washington, D.C., Amer. Meteor. Soc., 116—122.

Browning, P. A., 1991: The VDUC Interactive ComputerSystem at the National Severe Storms Forecast Center.Proc., Seventh Int. Conf on Interactive Informationand Processing Systems for Meteor., Oceanogr., andHydrol., New Orleans, Amer. Meteor. Soc., 204—207.

____, 1992: Use of interactive workstations in the NSSFCforecast operations. Preprints, Eighth InternationalConf Interactive Info. and Processing Systems forMeteor.. Oceanogr., and Hydrol.. Atlanta, Amer.Meteor. Soc., 233—237.

Burgess, D. W., and P. 5. Ray, 1986: Principles of radar.Mesoscale Meteorology and Forecasting, P. 5. Ray,Ed., Amer. Meteor. Soc., 85—117.

____, and L. R. Lemon, 1990: Severe thunderstormdetection by radar. Radar in Meteorology, D. Atlas,

Ed., Amer. Meteor. Soc., 619—647.____, and B. F. Smull, 1990: Doppler radar observations

of a bow echo associated with a long-track severewindstorm. Preprints, 16th Conf Severe Local Storms,Kananaskis Park. Alberta, Canada, Amer. Meteor. Soc.,203—208.

Caracena, F., J. McCarthy, and 1. A. Flueck, 1983:Forecasting the likelihood of microbursts along thefront range of Colorado. Preprints, 13th Conf SevereLocal Storms, Tulsa, Oklahoma, Amer. Meteor. Soc.,261—264.

Carr, F. H., and J. P. Millard, 1985: A composite study ofcomma clouds and the association with severe weatherover the Great Plains. Mon. Wea. Rev., 113, 370—387.

Charba, J., and Y. Sasaki, 1971: Structure and movementof the severe thunderstorms of 3 April 1964 as revealedfrom radar and surface mesonetwork data analysis. J.Meteor. Soc. Japan. 49, 191—213.

Chisholm, A. J., and J. H. Renick, 1972: The kinematicsof multicell and supercell Alberta hailstorms. AlbertaHail Studies— 1972, Research Council of Alberta HailStudies, Rep. 72-2, Edmonton, Canada, 24—31.

Colquhoun, J. R., 1987: A decision tree method offorecasting thunderstorms, severe thunderstorms andtornadoes. Wea. Forecasting, 2, 337—345.

Cooley, J. R., 1978: Cold air funnel clouds. Mon. Wea.Rev., 106, 1368—1372.

Cope, A. M., 1992: Visualization of numerical modeloutput at the National Severe Storms Forecast Center.Preprints, 8th International Conf Interactive Info, andProcessing Systems .for Meteor., Oceanogr., andHydrol., Atlanta, Amer. Meteor. Soc., 141 —144.

Crawford, C. C.. F. V. Brock, R. L. Elliot, G. W. Cuperus,S. J. Stadler, H. L. Johnson, and C. A. Doswell III,1992: The Oklahoma mesonetwork—A 21st centuryproject. Preprints, Eighth International ConfInteractive Info, and Processing Systems/or Meteor.,Oceanogr., and Hydrol., Atlanta, Amer. Meteor. Soc.,27—33.

Darkow, G. L., and D. W. McCann, 1977: Relativeenvironmental winds for 121 tornado bearing storms.Preprints, 11th Conf Severe Local Storms, Omaha,Nebraska, Amer. Meteor. Soc., 413—417.

Davies, J. M., 1989: On the use of shear magnitudes andhodographs in tornado forecasting. Preprints, 12thConf Wea. Analysis and Forecasting, Monterey,California, Amer. Meteor. Soc., 219—224.

____ and R. H. Johns, 1993: Some wind and instabilityparameters associated with strong and violenttornadoes. Part I: Wind shear and helicity. Proc.,Tornado Symposium III, C. Church, Ed., Amer.Geophys. Union, (in press).

DECEMBER 1992 J O H N S A N D D O S W E L L

Davies-Jones, R. P., 1984: Streamwise vorticity: The originof updraft rotation in supercell storms. J. Atmos. Sci.,41, 2991—3006.

____ and H. Brooks, 1993: Mesocyclogenesis from atheoretical perspective. Proc., Tornado Symp. III, C.Church, Ed., Amer. Geophys. Union, (in press).

____ D. Burgess, and M. Foster, 1990: Test of helicity asa tornado forecast parameter. Preprints, 16th ConfSevere Local Storms, Kananaskis Park, Alberta,Canada, Amer. Mete6r. Soc., 588—592.

Doswell, C. A. III, 1980: Synoptic-scale environmentsassociated with High Plains severe thunderstorms.Bull. Amer. Meteor. Soc., 61, 1388—1400.

____ 1982: The operational meteorology of convectiveweather. Vol. I: Operational Mesoanalysis. NOAATech. Memo NWS NSSFC5, 158 pp.

____, 1986a: The human element in weather forecasting.Natl. Wea. Dig., 11,2,6—17.

____, 1986b: Short-range forecasting. MesoscaleMeteorology and Forecasting, P. 5. Ray, Ed., Amer.Meteor. Soc., 689—719.

____, 1987: The distinction between large-scale andmesoscale contribution to severe convection: A casestudy example. Wea. Forecasting, 2, 3—16.

____ 1991a: A review for forecasters on the applicationsof hodographs to forecasting severe thunderstorms.Natl. Wea. Dig., 16, 1, 2—16.

____, 1991b: Comments on “Mesoscale convectivepatterns of the southern High Plains.” Bull. Amer.Meteor. Soc., 72, 389—390.

____, and D. W. Burgess, 1993: Tornadoes and tornadicstorms: A review of conceptual models. Proc., TornadoSymp. III, C. Church. Ed., Amer. Geophys. Union, (inpress).

____ J. T. Schaefer, D. W. McCann, T. W. Seblatter, andH. B. Wobus, 1982: Thermodynamic analysisprocedures at the National Severe Storms ForecastCenter. Preprints, Ninth Conf Wea. Forecasting andAnalysis, Seattle, Amer. Meteor. Soc., 304—309.

____, A. R. Moller, and R. Przybylinski, 1990: A unifiedset of conceptual models for variations on the supercelltheme. Preprints, 16th Conf Severe Local Storms,Kananaskis Park, Alberta, Canada, Amer. Meteor.Soc., 40—45.

____, S. J. Weiss, and R. H. Johns, 1993: Tornadoforecasting—a review. Proc., Tornado Symp. III, C.Church, Ed., Amer. Geophys. Union, (in press).

Dunn, L., 1986: An example of subjective interpretationof network profiler data in real-time forecasting. Wea.Forecasting, 1, 219—225.

Fawbush, E. J., and R. C. Miller, 1953: A method forforecasting hailstone size at the earth’s surface. Bull.Amer. Meteor. Soc., 34, 235—244.

____ and , 1954: A basis for forecasting peak wind gusts

in nonfrontal thunderstorms. Bull. Amer. Meteor. Soc.,35, 14—19.

Forbes, G. S., M. J. Markus, and G. D. Lessens, 1980:Some synoptic and mesoscale factors associated withdownburst-producing thunderstorms. Preprints, EighthConf Wea. Forecasting and Analysis, Denver, Amer.Meteor. Soc., 363—369.

Foster, D. 5., 1958: Thunderstorm gusts compared withcomputed downdraft speeds. Mon. Wea. Rev., 86, 91–94.

____, and F. Bates, 1956: A hail size forecasting technique.Bull. Amer. Meteor. Soc., 37, 135—140.

Frisch, A. S., B. W. Orr, and B. E. Martner, 1992: Dopplerradar observations of the development of a boundary-layer nocturnal jet. Mon. Wea. Rev., 120, 3—16.

Fujita, T. T., 1971: Proposed characterization of tornadoesand hurricanes by area and intensity. SMRP Res. Pap.No. 91, Dept. of Geophysical Sciences, University ofChicago, 42 pp.

____, 1978: Manual of downburst identification for projectNIMROD. Satellite and Mesometeorology Res. Pap.No. 156, University of Chicago, Dept. of GeophysicalSciences, 104 pp.

____, and H. R. Byers, 1977: Spearhead echo anddownburst in the crash of an airliner. Mon. Wea. Rev.,105, 129—146.

____, and R. M. Wakimoto, 1981: Five scales of airflowassociated with a series of downbursts on 16 July 1980.Mon. Wea. Rev., 109, 1438—1456.

Galway, J. G., 1956: The lifted index as a predictor of latentinstability. Bull. Amer. Meteor. Soc., 37, 528—529.

____, 1977: Some climatological aspects of tornadooutbreaks. Mon. Wea. Rev., 105, 477—484.

____, and A. Pearson, 1981: Winter tornado outbreaks.Mon. Wea. Rev., 109, 1072—1080.

Giordano, L. A., and J. M. Fritsch, 1991: Strong tornadoesand flash- flood producing rainstorms during the warmseason in the Mid- Atlantic region. Wea. Forecasting,6, 437—455.

Goetsch, E. H., 1988: Forecasting cold core severe weatheroutbreaks. Preprints, 15th Conf Severe Local Storms,Baltimore, Amer. Meteor. Soc., 468—471.

Hales, J. E., Jr., 1979: A subjective assessment of modelinitial conditions using satellite imagery. Bull. Amer.Meteor. Soc., 60, 206—211.

____, 1985: Synoptic features associated with Los Angelestornado occurrences. Bull. Amer. Meteor. Soc., 66,657—662.

____, 1988: Improving the watch/warning programthrough use of significant events data. Preprints, 15thConf Severe Local Storms, Baltimore, Amer. Meteor.Soc.. 165—168.

____, and C. A. Doswell III, 1982: High resolutiondiagnosis of instability using hourly sufface lifted

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

parcel temperatures. Preprints, 12th Conf Severe LocalStorms, San Antonio, Texas, Amer. Meteor. Soc.,172—175.

Hart, J. A., and W. D. Korotky. 1991: The SHARPWorkstation— vISO. A skew T/hodograph analysisand research program for the IBM and compatible PC.User’s Manual. NOAA/NWS Forecast Office,Charleston, WV, 62 pp.

Hirt, W. D., 1985: Forecasting severe weather in NorthDakota. Preprints, 14th Conf Severe Local Storms,Indianapolis, Indiana, Amer. Meteor. Soc., 328—331.

Johns, R. H., 1982: A synoptic climatology of northwestflow severe weather outbreaks. Part 1: Nature andsignificance. Mon. Wea. Rev., 110, 1653—1663.

____, 1984: A synoptic climatology of northwest flowsevere weather outbreaks. Part II: Meteorologicalparameters and synoptic patterns. Mon. Wea. Rev., 112,449—464.

____, 1993: Meteorological conditions associated withbow-echo development. Wea. Forecasting, (in press).

____, and W. D. Hirt, 1987: Derechos: Widespreadconvectively induced windstorms. Wea. Forecasting,2, 32—49.

____, and P. W. Leftwich, Jr., 1988: The severethunderstorm outbreak of July 28—29 1986. A caseexhibiting both isolated supercells and a derechoproducing convective system. Preprints, 15th ConfSevere Local Storms, Baltimore, Amer. Meteor. Soc.,448—451.

____,and W. R. Sammler, 1989: A preliminary synopticclimatology of violent tornado outbreaks utilizingradiosonde standard level data. Preprints, 12th ConfWea. Analysis and Forecasting, Monterey, California,Amer. Meteor. Soc., 196—201.

____, K. W. Howard, and R. A. Maddox, I 990a: Conditionsassociated with long-lived derechos—An examinationof the large scale environment. Preprints, 16th ConfSevere Local Storms, Kananaskis Park, Alberta,Canada, Amer. Meteor. Soc., 408—412.

____, J. M. Davies, and P. W. Leftwich, I 990b: Anexamination of the relationship of 0–2 km AGL“positive’ wind shear to potential buoyant energy instrong and violent tornado situations. Preprints, 16thConf Severe Local Storms, Kananaskis Park, Alberta,Canada, Amer. Meteor. Soc., 593—598.

____, ____, and____, 1993: Some wind and instabilityparameters associated with strong and violenttornadoes. Part II: Variations in the combinations ofwind and instability parameters. Proc. TornadoSymposium III, C. Church, Ed., Amer. Geophys.Union, (in press).

July, M. J., 1990: Forcing factors in the violent tornadooutbreak of May 5, 1989: A study of scale interaction.Preprints, 16th Conf Severe Local Storms, Kananaskis

Park, Alberta, Canada, Amer. Meteor. Soc., 72—77.Junker, N. W., J. E. Hoke, and R. H. Grumm, 1989:

Performance of NMC’s regional models. Wea.Forecasting, 4, 368—390.

Kamburova, P. L., and F. H. Ludlam, 1966: Rainfallevaporation in thunderstorm downdraughts. Quart. IRoy. Meteor. Soc., 92, 510—518.

Kelly, D. L., J. T. Schaefer, R. P. MeNulty, C. A. DoswellIII, and R. F. Abbey, Jr., 1978: An augmented tornadoclimatology. Mon. Wea. Rev., 106, 1172—1183.

____ and C. A. Doswell III, 1985: The climatology ofnontornadic severe thunderstorm events in the UnitedStates. Mon. Wea. Rev., 113, 1997—2014.

Knight, C. A., and P. Squires, Eds., 1981: Hailstorms ofthe Central High Plains. Vol. 1: The National HailResearch Experiment. Colorado Associated UniversityPress, 282 pp.

Korotky. W., R. W. Przybylinski, and J. A. Hart, 1993:The Plainfield, Illinois tornado of August 28, 1990:The evolution of synoptic and mesoscaleenvironments. Proc., Tornado Symposium III, C.Church, Ed., Amer. Geophys. Union, (in press).

Krumm, W. R., 1954: On the cause of downdrafts fromdry thunderstorms over the plateau area of the UnitedStates. Bull. Amer. Meteor. Soc., 35, 122—126.

Lanicci, J. M., and T. T. Warner, 1991: A synopticclimatology of the elevated mixed-layer inversion overthe southern Great Plains in spring. Part I: Structure,dynamics, and seasonal evolution. Wea. Forecasting,6, 181—197.

LaPenta, K. D., R. J. Kane, and J. S. Waldstreicher, 1990:A multi- scale examination of the 10 July 1989northeast tornado outbreak. Preprints, 16th Conf.Severe Local Storms, Kananaskis Park, Alberta,Canada, Amer. Meteor. Soc., 548—553.

Lazarus, S. M., and K. K. Droegemeier, 1990: Theinfluence of helicity on the stability and morphologyof numerically simulated storms. Preprints, 16th ConfSevere Local Storms, Kananaskis Park, Alberta,Canada, Amer. Meteor. Soc., 269—274.

Leftwich, P. W., 1984: Operational experiments inprediction of maximum expected hailstone diameter.Preprints, 10th Conf Wea. Forecasting and Analysis,Clearwater Beach, Florida, Amer. Meteor. Soc., 525–528.

____, and S. K. Beckman, 1992: A preliminary assessmentof the use of 404 MHz wind profiler data at theNational Severe Storms Forecast Center, J. Atmos.Oceanic Technol., (in press).

Lemon, L. R., and C. A. Doswell III, 1979: Severethunderstorm evolution and mesocyclone structure asrelated to tornadogenesis. Mon. Wea. Rev., 107, 1184—1197.

Lewis, J., 1989: Real time lightning data and its application

DECEMBER 1992 J O H N S A N D D O S W E L L

in forecasting convective activity. Preprints, 12th ConfWea. Analysis and Forecasting, Monterey, California,Amer. Meteor. Soc., 97—102.

Lilly, D. K., 1986: The structure, energetics, andpropagation of rotating convective storms. Part II:Helicity and storm stabilization. J. Atmos. Sci.. 43,126—140.

Maddox, R. A., 1976: An evaluation of tornado proximitywind and stability data. Mon. Wea. Rev., 104, 133—142.

____ 1993: The diurnal low-level wind oscillation andstorm-relative helicity. Proc., Tornado Symp. III, C.Church, Ed., Amer. Geophys. Union, (in press).

____, and C. A. Doswell III, 1982: An examination ofjetstream configurations, 500 mb vorticity advectionand low-level thermal advection patterns duringextended periods of intense convection. Mon. Wea.Rev., 110, 184—197.

____, L. R. Hoxit, and C. F. Chappell, 1980: A study oftornadic thunderstorm interactions with thermalboundaries. Mon. Wea. Rev., 108, 322—336.

McCaul, E. W., Jr., 1990: Simulations of convective stormsin hurricane environments. Preprints, 16th Conf SevereLocal Storms,Kananaskis Park, Alberta, Canada,Amer. Meteor. Soc., 334—339.

____, 1991: Buoyancy and shear characteristics ofhurricane—tornado environments. Mon. Wea. Rev.,119, 1954—1978.

MeNulty, R. P., 1978: On upper kinematics and severeweather occurrence. Mon. Wea. Rev., 106, 662—672.

____, 1981: Tornadoes west of the divide: A climatology.Natl. Wea. Dig., 6, 2, 26—30.

Miller, R. C., 1972: Notes on analysis and severe stormsforecasting procedures of the Air Force GlobalWeather Central. Tech. Rep. 200 (rev.), Air WeatherService, 181 pp.

Moller, A. R., C. A. Doswell, III, and R. Przybylinski, 1990:High precipitation supercells: A conceptual model anddocumentation. Preprints, 16th Conf Severe LocalStorms, Kananaskis Park, Alberta, Canada, Amer.Meteor. Soc., 52—57.

Moore, J. T., and H. A. Elkins, 1985: A synoptic analysisof the 6— 7 May 1975 Omaha tornado outbreak. Natl.Wea. Dig., 10, 1, 39—56.

____, and J. P. Pino, 1990: An interactive method forestimating maximum hailstone size from forecastsoundings. Wea. Forecasting. 5, 508—525.

Morgan, G. M., Jr., and P. W. Summers, 1986: Hailfalland hailstorm characteristics. Thunderstorms: ASocial, Scientific, and Technological Documentary.Vol. 2: Thunderstorm Morphology and Dynamics, 2nded., E. Kessler, Ed., University of Oklahoma Press,237—257.

Nelson, S. P., 1983: The influence of storm flow structure

on hail growth. J. Atmos. Sci.. 40, 1965—1983.Nolen, R. H., 1959: A radar pattern associated with

tornadoes. Bull. Amer. Meteor. Soc., 40, 277—279.Ostby, F. P., 1992: Operations of the National Severe

Storms Forecast Center. Wea. Forecasting, 7, 546—563.

Piltz, S. F., 1992: Operational hourly helicity—Anexperiment. Southern Topics Technical AttachmentSR/SSD 92-2, Southern Region Headquarters NWS,Ft. Worth, TX, 5 pp.

Pino, J. P., and J. T. Moore, 1989: An interactive methodfor estimating hailstone size and convectively-drivenwind gusts from forecast soundings. Preprints, 12thConf Wea. Analysis and Forecasting, Monterey,California, Amer. Meteor. Soc., 103—106.

Plummer, D. W., 1989: Diagnostic and forecast graphicsproducts at NMC using high frequency model output.Wea. Forecasting, 4, 83—89.

Prosser, N. E., and D. S. Foster, 1966: Upper air soundinganalysis by use of an electronic computer. J. Appl.Meteor., 5, 296—300.

Przybylinski, R. W., 1988: Radar signatures associated withthe 10 March 1986 tornado outbreak over centralIndiana. Preprints, 15th Conf Severe Local Storms.Baltimore, Amer. Meteor. Soc., 253—256.

____, and D. M. DeCaire, 1985: Radar signaturesassociated with the derecho, a type of mesoscaleconvective system. Preprints, 14th Conf Severe LocalStorms. Indianapolis, Indiana, Amer. Meteor. Soc.,228—231.

____, S. Runnels, P. Spoden, and S. Summy, 1990: TheAllendale, Illinois tornado—January 7, 1989. One typeof an HP supercell. Preprints, 16th Conf Severe LocalStorms, Kananaskis Park, Alberta, Canada, Amer.Meteor. Soc., 516—521.

Rasmussen, E. N., and R. B. Wilhelmson, 1983:Relationships between storm characteristics and 1200GMT hodographs, low- level shear, and stability.Preprints, 13th Conf Severe Local Storms, Tulsa,Oklahoma, Amer. Meteor. Soc., J5—J8.

Read, W. L.. 1987: Observed microbursts in the NWSSouthern Region during 1986—Four case studies.NOAA Tech. Memo. NWS SR-121. 10—34.

Rotunno, R., and J. B. Klemp, 1982: The influence of shear-induced vertical pressure gradient on thunderstormmotion. Mon. Wea. Rev., 110, 136—151.

____, and , 1985: On the rotation and propagation ofsimulated supercell thunderstorms. J. Atmos. Sci., 42,271—292.

Schaefer, J. T., 1986: Severe thunderstorm forecasting: Ahistorical perspective. Wea. Forecasting, 1, 164—189.

Schmidt, J. M., and W. R. Cotton, 1989: A High Plainssquall line associated with severe surface winds. J.Atmos. Sci., 46, 281—302.

W E A T H E R A N D F O R E C A S T I N G VOLUME 7

____, C. J. Tremback, and W. R. Cotton, 1990: Numericalsimulations of a derecho event: Synoptic andmesoscale components. Preprints, 16th Conf SevereLocal Storms, Kananaskis Park, Alberta, Canada,Amer. Meteor. Soc., 422—427.

Scofield, R and J. F. W. Purdom, 1986: The use of satellitedata for mesoscale analyses and forecastingapplications. Mesoscale Meteorology and Forecasting,P. 5. Ray, Ed., Amer. Meteor. Soc., 118—150.

Smith, B. E., 1990: Mesoscale structure of a derecho-producing convective system: The southern GreatPlains storms of May 4, 1989. Preprints, 16th ConfSevere Local Storms, Kananaskis Park, Alberta,Canada, Amer. Meteor. Soc., 455—460.

____, and J. W. Partacz, 1985: Bow echo induced tornadoat Minneapolis on 26 April 1984. Preprints, 14th ConfSevere Local Storms, Indianapolis, Indiana, Amer.Meteor. Soc., 81—84.

Smull, B. F., and R. A. Houze, Jr., 1987: Rear inflow insquall lines with trailing stratiform precipitation. Mon.Wea. Rev., 115. 2869—2889.

Sohl, C. J., 1987: Observed microbursts in the NWSSouthern Region during 1986—Four case studies.NOAA Tech. Memo. NWS SR-121, 1—9.

Srivastava, R. C., 1985: A simple model of evaporativelydriven downdraft: Application to a microburstdowndraft. J. Atmos. Sci., 42, 1004—1023.

Stone, H. M., 1988: Convection parameters and hodographprogram-CONVECTA and CONVECTB. NOAATech. Memo. NWS ERCP No.37 (revised), EasternRegion Headquarters, Bohemia, NY, 44 pp.

Szoke, E. J., and J. A. Augustine, 1990: A decade of tornadooccurrence associated with a surface mesoscale flowfeature—The Denver cyclone. Preprints, 16th ConfSevere Local Storms, Kananaskis Park, Alberta,Canada, Amer. Meteor. Soc., 554—559.

Turcotte, V., and D. Vigneux, 1987: Severe thunderstormand hail forecasting using derived parameters fromstandard RAOB data. Preprints, Second Workshop onOperational Meteor., Halifax, Nova Scotia, Canada,Atmos. Env. Serv./Canadian Meteor, and Oceanogr.Soc., 142—153.

Uccellini, L. W., 1990: The relationship between jet streaksand severe convective storm systems. Preprints, 16thConf Severe Local Storms, Kananaskis Park, Alberta,Canada. Amer. Meteor. Soc., 121—129.

____ and D. R. Johnson, 1979: The coupling of upper andlower tropospheric jet streaks and implications for thedevelopment of severe convective storms. Mon. Wea.Rev., 107, 682—703.

Wakimoto, R. M., 1985: Forecasting dry microburstactivity over the High Plains. Mon. Wea. Rev., 113,1131—1143.

____ and J. W. Wilson, 1989: Non-supercell tornadoes.

Mon. Wea. Rev., 117, 1113—1140.Weisman, M. L., 1990: The genesis of bow echoes: A rear-

inflow induced meso-convective structure.Cooperative Thesis No. 125, The Pennsylvania StateUniversity and National Center for AtmosphericResearch. NCARICT-125, 149 pp.

____ and J. B. Klemp, 1984: The structure andclassification of numerically simulated convectivestorms in directionally varying wind shears. Mon. Wea.Rev., 112, 2479—2498.

Weiss, S. J., 1987a: An assessment of the NGM four-layerlifted index prognoses of extreme instability. Natl.Wea. Dig., 12,2, 21—31.

____ 1987b: Some climatological aspects of forecastingtornadoes associated with tropical cyclones. Preprints,17th Conf Hurr. and Trop. Meteor., Miami, Amer.Meteor. Soc., 160—163.

Wilson, J. W., J. A. Moore, G. B. Foote, B. Martner, A. R.Rodi, T. Uttal, and J. M. Wilezak, 1988: ConvectionInitiation and Downburst Experiment (CINDE). Bull.Amer. Meteor. Soc., 69, 1328—1348.

Woodall, G. R., 1990: Qualitative forecasting of tornadicactivity using storm-relative environmental helicity.Preprints, 16th Conf Severe Local Storms, KananaskisPark, Alberta, Canada, Amer. Meteor. Soc., 311—315.